DLS 2020
Sun 15 - Fri 20 November 2020 Online Conference
co-located with SPLASH 2020
Wed 18 Nov 2020 17:00 - 17:20 at SPLASH-III - 5 Chair(s): Patrick Cousot, Sukyoung Ryu
Thu 19 Nov 2020 05:00 - 05:20 at SPLASH-III - 5 Chair(s): Xavier Rival, Sukyoung Ryu

Deep Neural Networks (DNNs) are rapidly being applied to safety-critical domains such as drone and airplane control, motivating techniques for verifying the safety of their behavior. Unfortunately, DNN verification is NP-hard, with current algorithms slowing exponentially with the number of nodes in the DNN. This paper introduces the notion of Abstract Neural Networks (ANNs), which can be used to soundly overapproximate DNNs while using fewer nodes. An ANN is like a DNN except weight matrices are replaced by values in a given abstract domain. We present a framework parameterized by the abstract domain and activation functions used in the DNN that can be used to construct a corresponding ANN. We present necessary and sufficient conditions on the DNN activation functions for the constructed ANN to soundly over-approximate the given DNN. Prior work on DNN abstraction was restricted to the interval domain and ReLU activation function. Our framework can be instantiated with other abstract domains such as octagons and polyhedra, as well as other activation functions such as Leaky ReLU, Sigmoid, and Hyperbolic Tangent.

Wed 18 Nov

Displayed time zone: Central Time (US & Canada) change

17:00 - 18:20
5DLS 2020 / SAS at SPLASH-III +12h
Chair(s): Patrick Cousot New York University, Sukyoung Ryu
17:00
20m
Research paper
Abstract Neural Networks
SAS
Matthew Sotoudeh University of California, Davis, Aditya V. Thakur University of California, Davis
Pre-print Media Attached
17:20
20m
Talk
Amalgamating Different JIT Compilations in a Meta-tracing JIT Compiler Framework
DLS 2020
Yusuke Izawa Tokyo Institute of Technology, Hidehiko Masuhara Tokyo Institute of Technology
Link to publication DOI Pre-print Media Attached
17:40
20m
Research paper
Probabilistic Lipschitz Analysis of Neural NetworksArtifact
SAS
Ravi Mangal Georgia Institute of Technology, Kartik Sarangmath Georgia Institute of Technology, Aditya Nori , Alessandro Orso Georgia Tech
Pre-print Media Attached
18:00
20m
Talk
Pricing Python Parallelism: A Dynamic Language Cost Model for Heterogeneous Platforms
DLS 2020
Dejice Jacob University of Glasgow, UK, Phil Trinder University of Glasgow, Jeremy Singer Glasgow University
Link to publication DOI Pre-print Media Attached

Thu 19 Nov

Displayed time zone: Central Time (US & Canada) change

05:00 - 06:20
5SAS / DLS 2020 at SPLASH-III
Chair(s): Xavier Rival INRIA/CNRS/ENS Paris, Sukyoung Ryu
05:00
20m
Research paper
Abstract Neural Networks
SAS
Matthew Sotoudeh University of California, Davis, Aditya V. Thakur University of California, Davis
Pre-print Media Attached
05:20
20m
Talk
Amalgamating Different JIT Compilations in a Meta-tracing JIT Compiler Framework
DLS 2020
Yusuke Izawa Tokyo Institute of Technology, Hidehiko Masuhara Tokyo Institute of Technology
Link to publication DOI Pre-print Media Attached
05:40
20m
Research paper
Probabilistic Lipschitz Analysis of Neural NetworksArtifact
SAS
Ravi Mangal Georgia Institute of Technology, Kartik Sarangmath Georgia Institute of Technology, Aditya Nori , Alessandro Orso Georgia Tech
Pre-print Media Attached
06:00
20m
Talk
Pricing Python Parallelism: A Dynamic Language Cost Model for Heterogeneous Platforms
DLS 2020
Dejice Jacob University of Glasgow, UK, Phil Trinder University of Glasgow, Jeremy Singer Glasgow University
Link to publication DOI Pre-print Media Attached